rdtand/Laguna-S-2.1-prismaquant-gridbook-6bit-vllm
877
serving section: gridbook 0.1.0 on PyPI (pip install gridbook), requirements table, corrected memory footprint
card metadata: library_name=gridbook, pipeline_tag, base_model_relation, expanded tags (discovery only; no install claims changed)
R6 LUT-resident decode: 2,186 tok/s prefill
auto-mode prefill numbers (plugin 2026-07-27)
A/B protocol note: match the measured gridbook config
performance section: consistent measured numbers (plugin 2026-07-23), config tradeoff, DFlash verdict
prefill 6.2-8.8x: updated A/B table (plugin 2026-07-23)
measured NVFP4 A/B table
Laguna-S-2.1 gridbook 6.0bpp artifact
model card
initial commit
